Sentence-level AI-generated text detection on MOSAIC XK
91.5Macro-F1SenFlow
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| SenFlowType=TB, Evaluation Protocol=Unified2026.06 | 91.5 | 0.968 | 0.831 | |
| SenDetEXType=TB, Evaluation Protocol=Unified2026.06 | 89.3 | 0.955 | 0.787 | |
| SeqXGPTType=TB, Evaluation Protocol=Unified2026.06 | 75.8 | 0.866 | 0.517 | |
| POGERType=TB, Evaluation Protocol=Unified2026.06 | 72.9 | 0.83 | 0.465 | |
| BinocularsType=TF, Evaluation Protocol=Unified, Threshold tuning strategy=single threshold tuned on the combined validation set2026.06 | 65.1 | 0.713 | 0.303 | |
| Fast-DetectGPTType=TF, Evaluation Protocol=Unified, Threshold tuning strategy=single threshold tuned on the combined validation set2026.06 | 57.7 | 0.59 | 0.153 |